Dan Riley

Hadn't thought of that. Will also note (one of my hobby horses) that "std::pow(x, 1.f/3.f);" is very different from "std::pow(x, 1./3.);". The former jumps straight to powf(), while the latter...

Tying back to my previous comment, "std::pow(x,2.f)+std::pow(y,2.f);" avoids the type promotion, and compiles down to two instructions if FMA is available. Fun, wasn't expecting that either.

> @dan131riley yeah, we shall just teach people to use 2.f ... Oh, I agree, teaching average physicists these corner cases is a non-starter. I mostly have the MkFit developers...

I thought we had changed the build params for eigen so it wouldn't use the `handmade_aligned_free()`. This is probably the same kind of alignment disagreement we were seeing before, I...

My understanding is that the normal mode of operations for the HLT nodes is to bind each cmsRun process to one socket and one GPU. Also, +1 to enabling `CUDAMonitoringService`.

> In principle we should expect the same results with `-O2` and `-O3` ? One could hope. At `-O3` the compiler will be more aggressive about taking advantage of undefined...

> I'm not sure if this crash could be related to #43802 I would guess that it is related to but perhaps not identical. In particular, #43802 does not seem...

el8_aarch64_gcc12, CMSSW_14_1_ROOT6_X_2024-05-15-2300, WF 11024.2 step3. Does this include the latest patches from @pcanal ? Clearer stack trace, no missing intermediates, and very likely related to #43802. ``` Thread 5 (Thread...

> Written that, it seems to me the frequency of problems with this package is increasing, and the package (in GitHub) seems practically unmaintained since 5 years, so I'm getting...

I suppose https://searchfox.org ought to be in consideration too?